Identification of indole glucosinolate breakdown products with antifeedant effects on Myzus persicae (green peach aphid)

Authors

  • Jae Hak Kim,

    1. Boyce Thompson Institute for Plant Research, Ithaca, NY 14853, USA
    Search for more papers by this author
    • Present address: Monsanto Company, 800 N. Lindbergh Blvd., St Louis, MO 63167, USA.

  • Byong Won Lee,

    1. Department of Chemistry and Chemical Biology, Cornell University, Ithaca, NY 148531301, USA
    Search for more papers by this author
    • Present address: Institute of Natural Medicine, Hallym University, Okchon-dong 1, Chunchon, Kangwon-do, 200-702, South Korea.

  • Frank C. Schroeder,

    1. Boyce Thompson Institute for Plant Research, Ithaca, NY 14853, USA
    2. Department of Chemistry and Chemical Biology, Cornell University, Ithaca, NY 148531301, USA
    Search for more papers by this author
  • Georg Jander

    Corresponding author
    1. Boyce Thompson Institute for Plant Research, Ithaca, NY 14853, USA
      (fax +1 607 254 1502; e-mail gj32@cornell.edu).
    Search for more papers by this author

(fax +1 607 254 1502; e-mail gj32@cornell.edu).

Summary

The cleavage of glucosinolates by myrosinase to produce toxic breakdown products is a characteristic insect defense of cruciferous plants. Although green peach aphids (Myzus persicae) are able to avoid most contact with myrosinase when feeding from the phloem of Arabidopsis thaliana, indole glucosinolates are nevertheless degraded during passage through the insects. A defensive role for indole glucosinolates is suggested by the observation that atr1D mutant plants, which overproduce indole glucosinolates, are more resistant to M. persicae, whereas cyp79B2 cyp79B3 double mutants, which lack indole glucosinolates, succumb to M. persicae more rapidly. Indole glucosinolate breakdown products, including conjugates formed with ascorbate, glutathione and amino acids, are elevated in the honeydew of M. persicae feeding from atr1D mutant plants, but are absent when the aphids are feeding on cyp79B2 cyp79B3 double mutants. M. persicae feeding from wild-type plants and myrosinase-deficient tgg1 tgg2 double mutants excrete a similar profile of indole glucosinolate-derived metabolites, indicating that the breakdown is independent of these foliar myrosinases. Artificial diet experiments show that the reaction of indole-3-carbinol, a breakdown product of indol-3-ylmethylglucosinolate, with ascorbate, glutathione and cysteine produces diindolylmethylcysteines and other conjugates that have antifeedant effects on M. persicae. Therefore, the post-ingestive breakdown of indole glucosinolates provides a defense against herbivores such as aphids that can avoid glucosinolate activation by plant myrosinases.

Introduction

Glucosinolates are defense-related secondary metabolites that are found almost exclusively in the Capparales. In the course of herbivory or other tissue damage, glucosinolates are activated through cleavage of a thioglucoside bond by endogenous plant myrosinases (β-thioglucosidases, EC3.2.1.147). Further breakdown results in the formation of thiocyanates, isothiocyanates, nitriles, epithionitriles, oxazolidine-2-thiones, and other potentially toxic or deterrent products (Halkier and Gershenzon, 2006). More than 120 glucosinolates have been identified in nature (Fahey et al., 2001), and there are likely to be several hundred possible glucosinolate breakdown products. Although this diversity may be required for defense against the large numbers of herbivores and pathogens that can attack cruciferous plants, so far very little is known about the defensive specificity of individual glucosinolates and their breakdown products.

The best-studied glucosinolate breakdown pathway is that of methionine-derived aliphatic glucosinolates in Arabidopsis thaliana and Brassica napus. The formation of nitriles rather than isothiocyanates is promoted by an epithiospecifier protein, which is present in some A. thaliana land races (Bernardi et al., 2000; Foo et al., 2000; Lambrix et al., 2001). More recently, an epithiomodifier protein, which promotes isothiocyanate formation, was found through the analysis of natural variation in A. thaliana glucosinolate breakdown (Zhang et al., 2006). Isothiocyanates are more toxic than nitriles for the larvae of Trichoplusia ni (cabbage looper; Lambrix et al., 2001) and Pieris rapae (cabbage butterfly; Wittstock et al., 2004), and for other insects.

During the breakdown of indol-3-ylmethylglucosinolate (I3M-GS), which is the predominant indole glucosinolate in A. thaliana and most Brassica species, the unstable initial products react rapidly to form indole-3-acetonitrile (IAN) and indole-3-carbinol (I3C, Figure 1). The epithiospecifier protein in A. thaliana promotes IAN formation in an analogous manner to the production of nitrile breakdown products from aliphatic glucosinolates (Burow et al., 2007). I3C is chemically less stable than IAN, and readily forms 3,3′-diindolylmethane and higher-order oligomers, both in purified form and in plant extracts (Buskov et al., 2000b). Indole-3-carboxaldehyde and methyl indole-3-carboxylate were the primary breakdown products of the isotopically labeled I3M-GS found in Brassica rapa (turnip) roots (Pedras et al., 2002). I3C, or the unstable thiohydroximate-O-sulfonate intermediate (Figure 1), reacts spontaneously with other plant metabolites, forming adducts with cysteine (indol-3-ylmethylcysteine; I3M-Cys), glutathione (indol-3-ylmethylglutathione; I3M-GSH) and ascorbate (ascorbigens) (Agerbirk et al., 1998; Buskov et al., 2000a; Staub et al., 2002). Initial breakdown of indole glucosinolates can occur even in the absence of myrosinase (Jürges and Thies, 1980), and non-enzymatic breakdown of I3M-GS in plant extracts has been reported (Tiedink et al., 1991).

Figure 1.

 Breakdown of indole-3-ylmethylglucosinolate (I3M-GS) after myrosinase cleavage.
Previously identified breakdown products and adducts formed with other plant compounds are shown.

A role for indole glucosinolates in crucifer defense is suggested by their induced accumulation in response to herbivory by root flies (Griffiths et al., 1994; Hopkins et al., 1998), lepidopteran larvae (Gigolashvili et al., 2007; Traw, 2002), flea beetles (Bartlet et al., 1999; Rostas et al., 2002), sawflies (Müller and Sieling, 2006) and aphids (Kim and Jander, 2007; Mewis et al., 2006). Moreover, several studies demostrate that jasmonate and other plant defense signaling molecules induce indole glucosinolate accumulation (Bartlet et al., 1999; Bodnaryk, 1994; Doughty et al., 1995; Kliebenstein et al., 2002; Mewis et al., 2005; Mikkelsen et al., 2004; Piotrowski et al., 2004; Sasaki-Sekimoto et al., 2005). Although Spodoptera exigua (beet armyworm), a generalist lepidopteran herbivore, avoids the high-indolic glucosinolate line hig1-1D in dual-choice assays (Gigolashvili et al., 2007), the insect-deterrent indole glucosinolate breakdown products that are presumed to cause this aversion have not yet been identified.

Green peach aphids (Myzus persicae) feed readily on A. thaliana and other crucifers. By inserting their stylets intercellularly to reach the phloem sieve elements (Tjallingii and Hogen Esch, 1993), the aphids may avoid bringing glucosinolate-containing phloem sap into contact with myrosinase, which is expressed in the cells adjacent to the phloem (Andreasson et al., 2001; Husebye et al., 2002; Thangstad et al., 2004). Consistent with this hypothesis, M. persicae reproduction on A. thaliana is not significantly affected by the presence or absence of the TGG1 and TGG2 myrosinases (Barth and Jander, 2006). Compared with aliphatic glucosinolates, indole glucosinolates suffer significant breakdown during passage through M. persicae (Kim and Jander, 2007), suggesting that they may contribute to aphid deterrence. Uptake of an artificial diet by M. persicae is almost completely blocked by breakdown products from 1 mm I3M-GS, showing that these products are antifeedant, rather than having post-ingestive effects on aphid reproduction (Kim and Jander, 2007). Nevertheless, attempts to correlate M. persicae reproduction with indole glucosinolate levels in A. thaliana defense-related mutants, and different Brassica isolates, produce mixed results (Cole, 1997; Mewis et al., 2005; Weber et al., 1986), most likely because of other genetic or metabolic variation in the plants being analyzed.

As a more direct investigation of the role that indole glucosinolates and their breakdown products play in plant defense, we studied M. persicae reproduction on A. thaliana accession Columbia-0 (Col-0) wild-type plants, the Col-0 cyp79B2 cyp79B3 double mutant, which is almost completely devoid of indole glucosinolates because of a biosynthetic block (Zhao et al., 2002), and Col-0 atr1D, a dominant transcription factor mutant with greatly elevated indole glucosinolate content (Celenza et al., 2005). Furthermore, by means of insect bioassays with purified I3M-GS breakdown products, and analysis of I3M-GS breakdown during M. persicae feeding on A. thaliana, we demonstrate that several breakdown products, including previously unknown cysteine adducts, significantly reduce M. persicae reproduction.

Results

Indole glucosinolate levels affect aphid reproduction

An analysis of M. persicae reproduction on mutant lines was used to determine whether indole glucosinolates can provide a defensive benefit to A. thaliana. Whereas the atr1D mutation increased foliar I3M-GS accumulation from six- to 22-fold, depending on the growth stage, cyp79B2 cyp79B3 double mutants were almost completely devoid of indole glucosinolates (Figure S1; Zhao et al., 2002; Celenza et al., 2005). In whole-plant experiments, the atr1D mutation significantly decreased aphid reproduction at the rosette growth stage, but not on flowering-stage plants (Figure 2a). Although M. persicae reproduction on cyp79B2 cyp79B3 double mutants was not significantly different from that on Col-0 wild-type plants in a short-term experiment (Figure 2b), the mutant plants succumbed to M. persicae more rapidly than the wild-type plants (Figure 2c). However, in control experiments without aphids, cyp79B2 cyp79B3 and wild-type plants grew equally well (Figure S2).

Figure 2.

 Aphid reproduction on indole glucosinolate mutants.
Progeny produced by one adult aphid over 5 days on (a) wild-type Col-0 vs. atr1D and (b) wild-type Col-0 vs. cyp79B2 cyp79B3 plants. Mean aphid number (+SD) of n = 16–25 replications; *P < 0.05, Student’s t-test. (c) Result of feeding by one aphid and its progeny for 30 days on cyp79B2 cyp79B3 (white arrows) and wild-type plants (black arrows) grown together in the same pots.

Indole glucosinolates are reduced in aphid honeydew compared with aliphatic glucosinolates

The glucosinolate content of phloem exudates and aphid honeydew were analyzed to determine whether there is glucosinolate breakdown during aphid feeding. Whereas indole glucosinolates constituted approximately 10% (wild type), 10%, (tgg1 tgg2) and 90% (atr1D) of the total glucosinolate content in phloem exudates (Figure 3a), they were reduced relative to aliphatic glucosinolates in aphid honeydew, to approximately 5% (wild type), 5% (tgg1 tgg2) and 20% (atr1D) of the total glucosinolates (Figure 3b). Unlike in whole plants (Figure S1) and phloem exudates (Figure 3a), intact methoxyindole glucosinolates (1MI3M-GS and 4MI3M-GS) were almost undetectable in aphid honeydew (Figure 3b). As expected, there were no detectable indole glucosinolates in experiments with cyp79B2 cyp79B3 double mutant plants. A similar loss of indole glucosinolates during passage through aphids on tgg1 tgg2 and wild-type plants showed that TGG1 and TGG2 are not involved in the breakdown of intact indole glucosinolates during aphid feeding.

Figure 3.

 Glucosinolate abundance in phloem exudates and aphid honeydew.
(a) Glucosinolates in phloem exudates; each bar represents the average of three independent samples, each of which consisted of the combined phloem sap collected from 15–20 plants.
(b) Glucosinolates in the honeydew of aphids feeding on cyp79B2 cyp79B3, Col-0 wild-type, tgg1 tgg2 and atr1D plants; average of eight samples. Indole glucosinolates are shown below the dashed line, and aliphatic glucosinolates are shown above the dashed line. Two other independent runs of these experiments gave similar results.

Indol-3-ylmethylglucosinolate breakdown products are present in aphid honeydew

Given the evident breakdown of indole glucosinolates during aphid feeding (Figure 3), we analyzed the aphid honeydew to identify breakdown products (Figure 4). In addition to known I3M-ascorbate and I3M-GSH adducts, three previously unknown amino acid conjugates (I3M-Pro, I3M-Leu and I3M-Ile) were identified by comparison of their HPLC retention times and MS fragmentation patterns (Figure S3) to synthesized standards.

Figure 4.

 Indole glucosinolate breakdown products in aphid honeydew.
(a, b) LC/MS/MS (single reaction monitoring) was used for quantifying indol-3-ylmethyl (I3M) conjugates.
(c) Relative ion intensities of more volatile compounds were measured by GC-MS, and were normalized with wild-type plants assigned the value of 1 for each compound. Indole-3-acetonitrile (IAN), indole-3-ethanol (I3E) and indole-3-carboxaldehyde (I3CAld). Mean ± SD of n = 4 or 5 replications. Two other independent runs of this experiment gave similar results.

Previously unknown conjugates with two indolylmethyl groups attached to cysteine, N,S-di(indol-3-ylmethyl)cysteine (I3M-Cys-I3M) and S-[2-(indol-3-ylmethyl)indol-3-ylmethyl]cysteine (I3M-I3M-Cys), were verified by HPLC-MS (Figure 5) and NMR spectroscopy (Figure S5). An additional unknown I3M conjugate (m/z = 259) was found in the honeydew of aphids feeding on atr1D plants (Figure S4). The MS fragmentation pattern (m/z = 130, as a major fragment) and the retention time (12.4 min) during HPLC separation suggested conjugate of I3M and a compound with a molecular weight 129, but it was not possible to identify the chemical structure based on the MS data alone. Although the molecular mass is the same as diindolylmethane (Figure 1), this unknown compound has a different HPLC retention time. IAN, indole-3-ethanol and indole-3-carboxaldehyde in aphid honeydew (Figures 4c, S6) were measured separately by GC-MS.

Figure 5.

 Di(indolylmethyl)cysteine structure identification.
(a) N,S-di(indol-3-ylmethyl)cysteine (I3M-Cys-I3M), and (b) S-[2-(indol-3-ylmethyl)indol-3-ylmethyl]cysteine (I3M-I3M-Cys) were identified by HPLC-MS, MS/MS and NMR spectroscopy (Figure S8). The molecular ion (m/z = 380) and fragments consistent with indole (m/z = 130), indol-3-ylmethylcysteine (m/z = 251), and two indole groups (m/z = 259) are indicated.

Consistent with the higher indole glucosinolate levels (Figure 3), the concentrations of all detected I3M-GS breakdown products were elevated in the honeydew from aphids feeding on atr1D mutant plants (Figure 4). I3M conjugates and other indole compounds were equally abundant (Student’s t-test, P > 0.05) in the honeydew from aphids on wild-type plants and tgg1 tgg2 double mutants, demonstrating that TGG1 and TGG2 are not required for indole glucosinolate breakdown during aphid feeding. No indole glucosinolate breakdown products were detected in the honeydew from aphids on cyp79B2 cyp79B3 plants. Breakdown products derived from 1MI3M-GS or 4MI3M-GS were also not detected in any aphid honeydew experiments.

Glucosinolate breakdown occurs within M. persicae

Plant material and whole aphids were assayed to determine whether indole glucosinolate breakdown was occurring within the plants or the insects. Ascorbigen was the only detectable I3M-GS breakdown product in plant tissue, and it was only found in the atr1D mutant (Figure S7). Furthermore, three days of aphid feeding did not affect the abundance of ascorbigen in plant tissue. I3M-Cys was detected in whole-body extracts of aphids feeding on wild-type plants, and tgg1 tgg2 and atr1D mutants (Figure S7), but not in the plants themselves. Together, these results suggest that I3M-GS breakdown during M. persicae feeding on wild-type A. thaliana is initiated after glucosinolate uptake.

Aphids feeding from plants or an artificial diet repeatedly extrude saliva (Miles, 1999). Therefore, intact I3M-GS was added to an artificial diet to determine whether aphid saliva contributes to indole glucosinolate breakdown. Approximately 50 aphids were confined in a cup covered with a Parafilm sachet containing an aphid diet. Irrespective of whether aphids were present during the 3-day experiment, ascorbigen, I3M-Trp and I3M-Cys were detected at similar concentrations in the diet (Figure S8), suggesting that aphid saliva does not contribute to this low level of spontaneous I3M-GS breakdown in an artificial diet.

Indole-3-carbinol (I3C) and cysteine combine to reduce aphid reproduction on an artificial diet

Further experiments were designed to determine whether the identified I3M-GS breakdown products inhibit M. persicae reproduction on artificial diets. Among seven commercially available I3M-GS breakdown products, four of which were found in aphid honeydew (Figure 4), only I3C significantly reduced aphid feeding and reproduction at 1 mm concentration (Figure 6a). Interestingly, sequential elimination of individual amino acids from the diet showed that the antifeedant effects of I3C were cysteine dependent (Figure 6b). As cysteine could be a limiting amino acid for M. persicae, we considered whether the removal of cysteine from the diet by reaction with I3C would reduce aphid growth and reproduction. However, this scenario seems unlikely because (i) the presence of cysteine was required to make I3C aphid-repellent, (ii) cysteine in the aphid diet was in excess over the added I3C (3.3 versus 1 mm), (iii) addition of more cysteine did not increase aphid reproduction and (iv) aphid reproduction was not significantly affected by the complete removal of cysteine from the diet (Figure 6b). The deleterious effects of I3C depended on the concentration, with an IC50 (the concentration of a compound needed to inhibit aphid reproduction on artificial diet by 50%) of 234 μm for I3C added to the standard 20-amino-acid aphid diet (Figure 6c).

Figure 6.

 Indole-3-carbinol (I3C) reduces aphid reproduction on artificial diets.
(a) Effect of seven indol-3-ylmethylglucosinolate (I3M-GS) breakdown products at 1 mM concentrations on aphid reproduction. The number of progeny for each compound is presented as a percentage of progeny on the control diet: mean (+SD; n = 10); **P < 0.01, Student’s t-test relative to diet-only control samples.
(b) The aphid-deterrent effects of 0.5 mm I3C, with and without cysteine in the diet. The number of progeny from four adult aphids after 3 days are shown: mean (+ SD) of n = 8 replications; **P < 0.01, Student’s t-test relative to diet-only control samples. S + 19AA: sucrose + 19 amino acids, without cysteine.
(c) Dose-response of aphid reproduction on artificial diet containing sucrose + 20 amino acids + I3C. Aphid numbers were normalized relative to reproduction on a control diet, which is set to 100%. Each data point in the graphs is the mean (±SD) of n = 10 replications at the given I3C concentration. The IC50 was calculated by a linear regression after a log transformation of the concentrations. The numbers in parentheses represent the confidence interval at α = 0.95.

To identify the presumed I3C-Cys adducts that reduce aphid reproduction, the aphid diet (I3C + sucrose + 20 amino acids) was separated by HPLC. Cysteine and tryptophan adducts were confirmed by MS and UV absorbance (at 280 nm; Figure S9), and by selectively removing amino acids from the diet prior to HPLC analysis (Figure S10). I3M-Ile, I3M-Leu and I3M-Pro adducts that were found in the honeydew of aphids feeding on A. thaliana (Figure 4) were not formed in the artificial diet assay. In aphid bioassays with fractions containing indole compounds, at concentrations equivalent to those when 1 mm I3C is added to the aphid diet, only I3M-Cys, I3M-I3M-Cys and I3M-Cys-I3M reduced aphid reproduction (Figure S9).

Structures of two abundant I3M-Trp isomers that formed in the artificial diet assay were confirmed by NMR spectroscopy, and are presented in Figure S11. These I3M-Trp adducts did not reduce aphid reproduction (Figure S9), nor did we observe these compounds in plant tissue or honeydew when aphids were feeding on A. thaliana. Therefore, it is unlikely that I3M-tryptophan adducts contribute significantly to defense against M. persicae.

The addition of I3C or myrosinase-catalyzed cleavage of I3M-GS in an aphid diet containing all 20 amino acids produce similar UV absorption profiles of indole-containing products (Figure 7a,b), except for the presence of ascorbigen, which results from the ascorbate found in the commercially available myrosinase preparation. This suggests that I3C, the predominant I3M-GS breakdown product under these conditions, is reacts immediately with other components of the aphid diet. Earlier research showed that the addition of myrosinase to I3M-GS causes a sixfold decrease in the IC50 for M. persicae reproduction (Kim and Jander, 2007). However, in contrast to the inhibitory effect of I3C itself (Figure 6b), the decrease in aphid reproduction caused by I3M-GS cleavage by myrosinase did not depend on the cysteine component of the aphid diet (Figure 7c). It is most likely that this is because of the antifeedant effects of ascorbigen (Figure 8f), which is found in the myrosinase-containing samples.

Figure 7.

 Indole compounds in the aphid diet with indole-3-carbinol (I3C) or indol-3-ylmethylglucosinolate (I3M-GS) + myrosinase.
HPLC-UV chromatograms at 280 nm of the 20-amino-acid aphid diet containing (a) 0.5 mm I3C or (b) 0.5 mm I3M + myrosinase. The identified major peaks in the chromatograms are: 1, S-indol-3-ylmethylcysteine; 2, ascorbigen; 3, I3C; 4, 5 and 7, I3M-Trp adducts; 6, di(indolylmethyl)cysteines; 8, diindolylmethane; 9, higher-order I3C adducts.
(c) Aphid reproduction on a diet with 0.2 mm I3M-GS and/or myrosinase: mean (+ SD) of progeny from four adult aphids after 3 days (n = 10); **P < 0.01, Student’s t-test comparison with control diet without I3M-GS or myrosinase. S + 18AA: sucrose + 18 amino acids, without cysteine and tryptophan.

Figure 8.

 Aphid reproduction on artificial diets containing indol-3-ylmethyl (I3M) conjugates.
Numbers of aphid progeny were normalized relative to reproduction on a control diet (sucrose + 18 amino acids, without cysteine and tryptophan), which is set to 100%. Each data point in the graphs is the mean (±SD) of n = 10 replications at that glucosinolate concentration. The IC50 was calculated by a linear regression after a log transformation of concentrations. The numbers in parentheses are confidence intervals at a = 0.95. GSH, glutathione.

I3M conjugates found in aphid honeydew inhibit aphid reproduction

I3M-Cys-I3M, I3M-I3M-Cys and I3M-Cys were purified by further HPLC fractionation. Aphid bioassays with purified compounds showed that I3M-I3M-Cys strongly reduced aphid reproduction, with an IC50 of 12 μm in artificial diet experiments (Figure 8a), whereas I3M-Cys-I3M has an IC50 of 209 μm (Figure 8b). The I3M-Cys adduct, which was described previously (Staub et al., 2002), has an IC50 of 245 μm (Figure 8c).

Other I3M conjugates were synthesized and confirmed by MS/MS and/or NMR spectroscopy. Ascorbigen, which was only detected in atr1D mutant plants (Figure S5), reduced aphid reproduction (IC50 = 269 μm; Figure 8f), an antifeedant activity that is similar to the action of I3M-Cys-I3M or I3M-Cys. Di(indol-3-ylmethyl)glutathione (I3M-I3M-GSH), which we only sporadically detected in the bodies and honeydew of aphids on A. thaliana, had an IC50 of 68 μm (Figure 8d), whereas the more abundant I3M-GSH had a comparatively moderate effect (IC50 = 437 μm, Figure 8e). The I3M-amino acid conjugates I3M-Pro, I3M-Ile and I3M-Leu did not reduce aphid reproduction at concentrations of up to 1 mm (Figure 8g–i).

Discussion

Together, our results support a model whereby TGG1- and TGG2-independent indole glucosinolate breakdown contributes to reduced M. persicae reproduction on A. thaliana: aphid reproduction is affected by the atr1D and cyp79B2 cyp79B3 indole glucosinolate mutations, indole glucosinolates are broken down during passage through the aphids feeding on A. thaliana, the abundance of the breakdown products correlated with the indole glucosinolate content of mutant and wild-type A. thaliana, and some of the identified breakdown products have significant antifeedant effects on M. persicae when added to artificial diets.

The atr1D mutation is associated with decreased reproduction on rosette-stage A. thaliana (Figure 2a), but not on flowering-stage plants, which have a lower glucosinolate content (Figure S1). M. persicae on flowering-stage A. thaliana feed almost exclusively from the flower stalks, which serve as a conduit for nutrients being transported to the seeds. Therefore, it is likely that glucosinolates and other factors, such as amino acids in the phloem sap, are integrated to determine the overall attractiveness of the plant, and that a positive stimulus from the abundant nutrients in the flower stalk phloem could outweigh the negative stimulus derived from indole glucosinolates in the atr1D mutant.

The cyp79B2 cyp79B3 double mutant plants succumbed more quickly to aphid infestation than the wild-type Col-0 plants. However, as aphid reproduction is also negatively affected by life on dead and dying plants, it was not possible to determine whether M. persicae reproduced faster on cyp79B2 cyp79B3 double mutants than on Col-0 plants in this experiment. Although differences in the indole glucosinolate abundance (Figures S1, 3) are likely to be the proximal cause of altered aphid sensitivity, we cannot completely rule out the effects of other downstream metabolites in cyp79B2 cyp79B3 double mutants. Accumulation of camalexin, a well-studied A. thaliana phytoalexin, is also abolished in cyp79B2 cyp79B3 double mutants (Glawischnig et al., 2004), but the pad3 mutation, which knocks out a cytochrome P450 catalyzing the final step in camalexin biosynthesis (Schuhegger et al., 2006), does not significantly affect M. persicae reproduction (Pegadaraju et al., 2005). A lack of camalexin could make plants more susceptible to pathogen attack (Ferrari et al., 2007; Glazebrook et al., 1997), and, in conjunction with aphid feeding, this might cause the early death of cyp79B2 cyp79B3 double mutant plants (Figure 2c). Another scenario that cannot be ruled is that the cyp79B2 cyp79B3 double mutants are somehow less nutritious than wild-type Col-0 plants, and that aphids, by compensating with increased consumption of phloem sap, cause the mutant line to die more quickly.

The formation of I3M-GS breakdown products during M. persicae feeding on A. thaliana occurs even in the absence of TGG1 and TGG2 (Figure 4). Several myrosinases have been detected through B. napus phloem sap proteomics (Giavalisco et al., 2006), and as yet unknown A. thaliana thioglucosidases may be active in the phloem during aphid feeding. Whether I3M-GS breakdown during aphid feeding is non-enzymatic, is catalyzed by plant-derived enzymes, or is the result of enzymes produced by the aphids themselves or their gut microflora, remains to be determined. The similar chemical profiles of no-aphid and aphid-fed artificial diets (Figure S8) show that aphid salivary enzymes are not involved in I3M-GS breakdown prior to ingestion. However, comparison of M. persicae cDNA sequences with other GenBank data identified a gene (accession number ES221351; Ramsey et al., 2007) that has greatest similarity to a well-studied myrosinase found in Brevicoryne brassicae (cabbage aphid; Francis et al., 2002; Jones et al., 2001; Pontoppidan et al., 2001). This suggests that myrosinase activity could be present in the gut or body of M. persicae, even if it is not found in the saliva.

Several I3M conjugates found in the honeydew of M. persicae feeding from A. thaliana reduced aphid reproduction in an artificial diet system (Figure 8), suggesting that they jointly contribute to the antifeedant effect of I3M-GS in A. thaliana (Figure 2). I3M-I3M-Cys (Figure 5) has a 17-fold lower IC50 than I3M-Cys-I3M (Figure 8a,b). As I3M-Cys was the predominant I3M conjugate detected in aphid bodies (Figure S7), aphid fitness might be improved if the insects could direct the further metabolism of I3M-Cys in their bodies towards I3M-Cys-I3M rather than I3M-I3M-Cys. Similarly, the formation of other I3M-amino acid conjugates that do not significantly inhibit aphid reproduction (I3M-Pro, I3M-Ile and I3M-Leu; Figure 8) could represent an M. persicae mechanism for detoxifying harmful indole glucosinolate breakdown products.

Jasmonic acid treatment of A. thaliana induces the formation of not only indole glucosinolates, but also of ascorbic acid, cysteine and glutathione (Sasaki-Sekimoto et al., 2005). Interestingly, all three of these upregulated compounds form antifeedant conjugates with I3C (Figure 8). The exogenous addition of jasmonate also reduces M. persicae reproduction, and jasmonate-insensitive coi1 mutants allow increased aphid reproduction (Ellis et al., 2002; Mewis et al., 2005). Therefore, upregulation of potential I3C conjugates, in conjunction with I3M-GS, may result in an elevated aphid resistance in plants that have defense responses induced via the jasmonate pathway.

I3M-I3M-Cys and other I3M-GS breakdown products may also contribute to plant defense against chewing herbivores, as is suggested by the deterrent effect of the hig1-1D mutation on S. exigua larvae (Gigolashvili et al., 2007). To date, the effect of indole glucosinolate breakdown products on chewing herbivores has not been confirmed. However, such experiments would be facilitated by the fact that several I3M-GS breakdown products are commercially available, as well as by the relatively simple preparation and purification of I3M adducts.

The methoxyindole glucosinolates 1MI3M-GS and 4MI3M-GS, even though they are less abundant in A. thaliana than I3M-GS, may also contribute significantly to aphid deterrence (Kim and Jander, 2007). The breakdown pathway of these secondarily modified indole glucosinolates is similar to that shown in Figure 1, but results in the formation of the corresponding methoxyindole products (Agerbirk et al., 1998; Bednarek et al., 2005; Hanley et al., 1985). In the absence of myrosinase, the I3M-GS IC50 for M. persicae reproduction is sixfold higher than that of 1MI3M-GS or 4MI3M-GS (Kim and Jander, 2007). A greater proportion of the methoxyindole glucosinolates than I3M-GS was lost during passage through M. persicae (Figure 3). If indole glucosinolates are differentially sensitive to enzymes in the aphid gut, the more rapid breakdown of 1MI3M-GS and 4MI3M-GS could cause these glucosinolates to have a stronger antifeedant effect than I3M-GS. Such variability in glucosinolate breakdown has been observed previously. For instance, 4MI3M-GS is degraded significantly less rapidly than 1MI3M-GS and I3M-GS by A. thaliana TGG1 and TGG2 (see Figure S1 in Barth and Jander, 2006).

I3M-I3M-Cys and I3M-I3M-GSH have a lower IC50 for aphid reproduction than I3M-Cys and I3M-GSH, respectively (Figure 8), suggesting that the two conjugated indolylmethyl groups somehow contribute to the antifeedant effect. Further research with M. persicae may determine the proximal cause of feeding deterrence, which has not been determined for these or any other glucosinolate breakdown products. The identification of antifeedant I3M-GS breakdown products, combined with reduced M. persicae reproduction on the atr1D mutant, also suggests practical applications, in that vegetable crops in the Brassica genus that are bred to have higher indole glucosinolate content might have greater resistance to aphids, and the viruses that they transmit.

In contrast to most prior research on the insect-deterrent effects of glucosinolate breakdown products, the experiments described here have involved indole, rather than aliphatic, glucosinolates. The wide distribution of indole glucosinolates among crucifers (Fahey et al., 2001; McDanell et al., 1988) suggests that they play a broad-spectrum role in deterring herbivory. Our experimental approach led to the identification of previously unknown indole glucosinolate breakdown products that reduce aphid reproduction. These findings will open up new lines of research for investigating the function of indole glucosinolates in defense against herbivores that consume cruciferous plants.

Experimental procedures

Plants, insects and growth conditions

Col-0 wild-type seeds were obtained from the Arabidopsis Biological Resource Center (ABRC, http://www.arabidopsis.org). The Col-0 cyp79B2 cyp79B3 double mutant (Zhao et al., 2002) was kindly supplied by J. Celenza (Boston University, Boston, MA, USA). Col-0 atr1D seeds (Bender and Fink, 1998; Celenza et al., 2005) were provided by J. Bender (Johns Hopkins University, Baltimore, MD, USA). The tgg1 tgg2 double knock-out mutant was described previously by Barth and Jander (2006). The growth conditions for plants and insects were as described previously by Kim and Jander (2007).

Aphid reproduction assays and glucosinolate analysis

One adult aphid was placed on a plant, the plant was covered with a mesh cup (Figure S2), and aphid progeny were counted after 5 days. To test whether aphid reproduction is dependent on the plant developmental stage, an adult aphid was placed on plants at four different times after sowing: days 10, 14 and 17 for the rosette-stage experiment, day 25 for wild-type plants, and tgg1 tgg2 and cyp79B2 cyp79B3 double mutants, and day 32 for atr1D plants, because atr1D mutant plants started flowering about 1 week later than Col-0 wild-type plants in our growth environment. Glucosinolate extraction and analysis were described previously by Kim and Jander (2007).

Extraction of indole compounds from honeydew, plants or aphid bodies

Flower stalks infested with aphids were used for honeydew collection. More than 50 aphids were placed on at least four flower stalks that crossed a Petri dish. The weight of total honeydew from aphid feeding was 10–30 mg per Petri dish. The Petri dish was washed with 5 ml of water followed by 2 ml of 60% methanol, after adding sinalbin (20 μl of 0.5 mm solution) and I3M-I3M-proline (2 μl of 2 mm solution) as internal standards for HPLC-MS. The volumes of methanol extract were reduced until ∼1 ml remained, using a Savant SC 110 rotary evaporator (Thermo Savant, –) and were then pooled with the water extract. For HPLC-MS analysis, a double-solid phase extraction method was developed to separate indole compounds from others. A column (96-well SPE VersaPlate Tubes; Varian, http://www.varian.com) was filled first with methanol-activated C18 resins (1-cm height; Chromtech, http://www.chromtech.com) and then water-activated Sephadex A-25 resin (1-cm height; Amersham Biosciences, http://www.amersham.com) was added. Concentrated honeydew extracts were applied onto the column and washed with water (3 ml). Sugars and most of amino acids except for tryptophan were eluted in this step. Indole and other hydrophobic compounds were washed off with 80% methanol (2 ml) and were then collected. The column was washed with 2 ml of water followed by a sulfatase treatment overnight for analyzing desulfoglucosinolates. The methanol extract was evaporated using a Savant SC 110 rotary evaporator, and redissolved in 60% methanol for storage at −80°C.

For GC-MS analysis, benzonitrile (20 μl of 8.3 mm solution) was added as an internal standard. Honeydew was washed with 60% methanol, and was then concentrated as described above. This concentrated extract was mixed with an equal volume of dichloromethane in an 8-ml glass vial and vortexed for 1 min. After centrifugation at 500 g for 10 min, the dichloromethane layer was filtered through a glass column containing glass wool and anhydrous Na2SO4. The remaining aqueous layer was re-extracted with another 1 ml of dichloromethane as described above, and the combined extracts were concentrated to approximately 200 μl under nitrogen before injecting into GC.

Aphid bodies (1 ml by volume) were collected in a microcentrifuge tube containing three 3-mm steel balls (Abbott Ball Company, http://www.abbottball.com) and were ground in 100% methanol by vortexing. After centrifugation at 13 000 g for 10 min, the supernatant was collected and the remaining aphid tissues were re-extracted with 80% methanol. Indole glucosinolate breakdown products and other compounds were separated using the solid-phase extraction described above.

Synthesis and purification of I3M conjugates

Indole-3-carbinol (50 mg) was mixed with an amino acid, glutathione or ascorbic acid in a 1:1 (w/w) ratio in a 20-ml glass vial, followed by adding 10 ml of 80% methanol. The vial was incubated at 85°C for 30 min in a water bath, vortexed for 5 min and was then left at 23°C overnight. The supernatant was collected after centrifugation (13 000 g for 15 min) and was directly injected (20 μl) into the HPLC with a LiChrospher RP18 column (Varian, http://www.varian.com) for the collection of the fraction. For HPLC separation, the mobile phases were: A, water, and B, 90% acetonitrile, at a flow rate of 1 ml min−1. The column linear gradients for samples were as follows: 0 min, 80% A; 0–10 min, 65% A; 10–12 min, 0%; 12–15 min, 0%; 15–16 min, 80%; 16–23 min, 80%. For a better separation of I3M-I3M-Cys and I3M-Cys-I3M, 0.1% formic acid as solvent A, and 90% acetonitrile with 0.1% formic acid as solvent B, at a flow rate of 1 ml min−1, were used. The linear gradient of solvent A was: 0 min, 90%; 0–20 min, 30%; 20–21 min, 0%; 21–26 min, 0%; 26–27 min, 90%; 27–35 min, 90%. The collected fractions were purified one more time using the first HPLC gradient system without formic acid in the solvent system. Chemical structures of purified I3M conjugates were confirmed with MS/MS. I3M-Cys, I3M-I3M-Cys and I3M-Cys-I3M were verified both using MS/MS and NMR spectroscopy. Standard curves for HPLC-purified I3M conjugates were drawn based on the area at 280 nm (UV) at four different concentrations.

Analysis of I3M conjugates (HPLC-MS, GC-MS and MS/MS)

All samples from aphid honeydew, plant extract, aphid bodies or synthesized standards were separated using a LiChrospher RP18 column (Varian) on a Waters Alliance 2695 HPLC system (Waters, http://www.waters.com), using the HPLC method described above for the separation of I3M-I3M-Cys from I3M-Cys-I3M. The eluent was monitored by a Waters 2996 Photodiode Array Detector and then 200 μl min−1 were infused into a Micromass Platform LC mass spectrometer (Waters) after a post-column splitting by a Micro-Splitter Valve (Upchurch Scientific, http://www.upchurch.com). The positive-ion full-scan mass spectra were recorded for a mass-to-charge (m/z) ratio from 100 to 600. The capillary and cone voltage in ESI (electrospray ionization) mode were 3.5 kV and 30 V, respectively, and the source heater was set at 100°C. I3M conjugates in samples were compared with I3M-conjugated standards by monitoring peak retention times, along with the corresponding molecular ions [M + H]+and their major fragment ions. The ion intensities of a selected molecular ion [M + H]+ in samples was used for normalization, with the wild type value set to 1.

The purified I3M conjugates dissolved in 50% acetonitrile (0.1% formic acid) were directly infused (10 μl of 200 nm solution per min) into a Varian 1200L quadrupole MS/MS (Varian) using a syringe pump (Harvard Apparatus, http://www.harvardapparatus.com). Both positive parent ion and daughter ion scan methods were used to confirm the chemical constituents. For the quantification of indole compounds in aphid honeydew, single-reaction monitoring (SRM) was applied, and the molecular ions and fragment ions used for MS/MS were as follows: I3M-I3M-Cys, m/z = 380 [M + H]+ and 259 [I3M-I3M + H]+; I3M-Cys-I3M, m/z = 380 [M + H]+ and 251 [I3M-Cys + H]+; I3M-Cys, m/z = 251 [M + H]+ and 130 [I3M + H]+; I3M-I3M-GSH, m/z = 566 [M + H]+ and 259 [I3M-I3M + H]+; I3M-GSH, m/z = 437 [M + H]+ and 308 [GSH + H]+; ascorbigen, m/z = 306 [M + H]+ and 130 [I3M + H]+; I3M-Pro, m/z = 251 [M + H]+ and 130 [I3M + H]+; I3M-Ile and I3M-Leu, m/z = 261 [M + H]+ and 130 [I3M + H]+; I3M-tryptophan, m/z = 334 [M + H]+ and 130 [I3M + H]+. The Waters 2790 HPLC system consisted of a Waters 2475 Multi λ Fluorescence detector and a Waters 2487 Dual λ Absorbance detector, and 0.1% formic acid as solvent A and 90% acetonitrile with 0.1% formic acid as solvent B at a flow rate of 0.2 ml min−1 were used. The linear gradient of solvent A was: 0 min, 85%; 0–5 min, 85%; 5–15 min, 50%; 15–20 min, 0%; 20–22 min, 0%; 22–23 min, 85%; 23–29 min, 85%. The Luna C18 column (3 micron particle size, 50 mm × 2.0 mm; Phenomenex, http://www.phenomenex.com) with a SecurityGuard Cartridge (4 × 2.0 mm; Phenomenex) was connected to the MS without a splitting. The MS conditions were as follows: needle, 5000 V; drying gas, 185°C, 19 psi; capillary scan, 10 V; shield, 600 V; detector, 1800 V on positive.

Dichloromethane extracts of aphid honeydew were used for the detection of IAN and other indole compounds by GC-MS (Varian CP-3800 GC/1200L quadrupole MS/MS) with a factorFOUR column (Varian, 30 m × 0.25 mm × 0.25 μm). The temperature was programmed as follows; 40°C for 5 min, 12°C min−1 to 300°C, with a 10-min final hold. Commercial IAN (m/z = 156), indole-3-ethanol (m/z = 161) and indole-3-carboxaldehyde (m/z = 145; Sigma-Aldrich, http://www.sigmaaldrich.com) were used to identify indole breakdown products in aphid honeydew by their retention times and molecular ions.

NMR spectroscopy

NMR spectra were recorded at 25°C by using a Varian INOVA (600-MHz proton, 151-MHz carbon) spectrometer with methanol-d4 as the solvent. (1H,1H)-double-quantum filtered correlation spectroscopy (dqf-COSY) spectra and phase-sensitive nuclear Overhauser effect spectroscopy (NOESY) spectra, as well as heteronuclear multiple-bond and multiple-quantum correlation spectra, were acquired by using the standard pulse sequences and phase cycling for coherence selection. For dqf-COSY spectra, an acquisition time of 0.6 s and 64 increments per ppm of sweep width were used. For NOESY spectra, a mixing time of 0.6 s, an acquisition time of 0.25 s and 32 increments per ppm of sweep width were used. Spectra were processed with MestReC or Varian vnmr software.

Artificial diet assays

Artificial diet assays and IC50 calculations were performed as described previously by Kim and Jander (2007). Four wingless adults were placed in a cup (30-ml Portion Cups; Comet Products, http://www.wna-inc.com), which was covered with a Parafilm sachet containing 45 μL of the liquid diet. After 3 days, aphid nymphs were counted. The IC50 was calculated based on the relative aphid reproduction at each concentration using a simple linear regression model (inverse prediction) using jmp v.6 (SAS Institute, http://www.sas.com) after a log transformation of diet concentrations.

Analysis of indole breakdown products in an artificial diet assay

Using the same experimental set-up as described above, with minor modifications, compounds mixed in artificial diets, with or without aphid feeding, and in honeydew were analyzed by HPLC-MS. Honeydew was collected on an aluminum foil that was placed on the bottom of the diet cup. The diet consisted of sucrose, 20 amino acids, sinigrin (0.5 mm), glutathione (0.5 mm), ascorbic acid (1 mm) and I3M-GS (0.15–0.35 mm). Approximately fifty aphids were caged and 50–100 μl of the diet was provided for 3 days. Three different collections from this experiment were compared: a diet without aphid in a Parafilm sachet, a diet left after aphid feeding and honeydew. Honeydew collected on aluminum foil was extracted using 60% methanol. The addition of myrosinase into an aphid diet containing I3M-GS and 20 amino acids to measure the production of indole breakdown products was described previously by Kim and Jander (2007).

Acknowledgements

We thank J. Bender and J. Celenza for providing the mutant A. thaliana lines, and M. Reichelt and J. Gershenzon for both technical advice and helpful comments regarding the manuscript. This research was funded by NSF grants OISE-0436554, DBI-0500550 and IOS-0718733, and USDA-CSREES grant 2005-35604-15446 to GJ, NIH grant GM53830 to FS, and Korea Research Foundation grant KRF-2005-F00028 to BL.

Ancillary